Mercedes-Benz Research & Development India Deploys Embedded AI for Cabin Comfort
AI-Powered Virtual Sensor Delivers Real-Time Cabin Air Mass Flow Estimation
“With the tooling solutions on Deep Learning Toolbox™, the overall development time has significantly [been] reduced. So, if you have to start off something from the concept phase to final target deployment, the overall duration is significantly reduced with the tool capabilities.”
Key Outcomes
- Reduced mean absolute error by 50% compared to physics-based models, while increasing RAM usage by just 1%
- Generated generic C/C++ code for 8-bit quantized deep neural networks
- Enabled onboard processing, ensuring compliance with data privacy regulations
Mercedes-Benz Research & Development India (MBRDI) focuses on developing innovative body and comfort features for next-generation vehicles. Accurate measurement or estimation of cabin air mass flow is critical for delivering personalized thermal comfort, regulating humidity, and maintaining air quality—factors that directly impact occupant well-being and energy efficiency. Traditional physics-based models often struggle to deliver the required accuracy and responsiveness for real-time cabin control. Alternatively, cloud-based AI models can provide improved accuracy, but they introduce challenges related to real-time execution, data transfer, and compliance with privacy regulations.
To overcome these limitations, the Body and Comfort team at MBRDI adopted an embedded AI approach, developing a deep learning–based virtual sensor for real-time mass flow estimation and deploying it locally to an ECU. Using MATLAB® and Simulink®, the team generated and preprocessed experimental data, designed and trained neural networks, compressed them with neural network projection and quantization, and deployed them by generating optimized production code with Embedded Coder®.
Products Used
Related Resources
